DynamoDB has it's applications but I don't like the pitch as a general database. It's basically: _yea but what if you want to scale your TODO application to the moon?? all it takes is just spewing inconsistent data everywhere and reinventing JOINs using string concatenation._
Very engaging presentation with useful techniques outlined clearly. It does pose the question of where the dividing line between a design pattern and a workaround for missing functionality lies.
I want to see how that actor/movie key works when you have something like Eddie Murphy and The Nutty Professor. Maybe that just wasn't a good example table, or I'm missing something? FYI, in that movie (like many other movies), the actor plays more than 1 role.
How does AWS Amplify handle data access patterns since it doesn’t ask the user to define them when the user is defining relationships? I assume it just sets up a lot of global secondary indexes which will consume way more WCUs and make it way more expensive to add to your data than if you set everything up manually
Maybe Mr Rick Houlihan continue helping our AWS Community. And I hope he takes a look how he introduces any topic or guest. Entonation and speed, a little better, less flat talking. Despite all, I really thank to all of you for every detail in this video. This is not a Daily News report but an updating event. Thanks.
Because the data is split up in 10G chunks across multiple partitions, you can't fetch data from other partitions in the same operation. A SQL join combines information from multiple tables matching a specific condition, which means that it has to search all of the columns for items that meet that condition. In order to maintain it's speed, DynamoDB opts to not allow operations like this. Sorry this is 2 months after you asked the question, but I hope my answer makes sense!
@@zanderkrasny7132 Even if multiple partitions are split across multiple computers, nosql could still (technically) offer joins by fetching across the different partitions no? Unless you are saying this would make it slow to point where it would be unusable and SQL solves this problem by having all tables on the same shard/partition/computer? From my understanding however the whole newSQL movement is about providing sharding to SQL databases. If they solve this problem, why can't noSQL databases also solve it?
@@samlaf92 From what I understand, it's not that they couldn't implement join operations across partitions, it's just that it would defeat the performance goals of partitioned noSQL. If you're searching every partition for what you're looking for, you lose the ability to run an O(1) hash, and then only searching 10G, because you have to go through all of the rest of your data, which can be massive. I don't know how newSQL gets around this; I haven't looked into it.
@@alexbdebrie kudos for checking comment section. don't lose your enthusiasm when you get to be the most seniorest dev in your company or in world or smth.